DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

CMR COLLEGE OF ENGINEERING AND TECHNOLOGY

. (An Autonomous institution, Approved by AICTE, Affiliated to JNTUH,NAAC’A+’)

Kandlakoya, Medchal Road, Hyderabad-501401

1
 A Project report on

HELEMET & NUMBER PLATED DETECTION OF A BIKE

A Dissertation submitted to JNTU Hyderabad in partial fulfilment of the academic
. requirements for the award of the degree.

Bachelor of Technology
In
Computer Science and Engineering
Submitted By
BANOTH GANESH (20H51A05D6)
KETHAVATH SURESH (21H55A0511)
VANAPARTHI KEERTHANA (21H55A0524)
Under the esteemed guidance of

Mr. V. Narasimha

Assistant Professor

Department of Computer science and Engineering

CMR COLLEGE OF ENGINEERING AND TECHNOLOGY

. (An Autonomous institution, Approved by AICTE, Affiliated to JNTUH,NAAC’A+’)
. Kandlakoya, Hyderabad-501401
2020-2024

2
 CMR COLLEGE OF ENGINEERING AND TECHNOLOGY
Kandlakoya, Medchal Road, Hyderabad-501401

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

CERTIFICATE
This is to certify that the Project report "Helmet & number plated detection of a Bike " being

submitted by BANOTH GANESH(20H51A05D6), KETHAVATH SURESH(21H55A0511),

VANAPARTHI KEERTHANA (21H55A0524), in partial fulfillment for the award of Bachelor of

Technology in Computer Science and Engineering of a record of bonafide work carried out his/her

under my guidance and supervision.

The results embodies in this project report have not been submitted any other

University or Institute for the award of any Degree.

Mr . V. Narasimha Dr. S. Siva Skandha

Assistant Professor Associate Professor
Dept .of CSE Head of dept.

3
 Acknowledgment
With great pleasure We want to take this opportunity to express my heartfelt gratitude to all

the people who helped in making this project work a grand success.

We are grateful to V. Narasimha ,Assistant Professor, Dept of Computer Science and

Engineering for his valuable technical suggestions and guidance during the execution of this

project work.

We would like to thank Dr. S. Siva Skandha, Associate Professor Head of the Department of

Computer Science and Engineering, CMR College of Engineering and Technology, who is the

major driving forces to complete my project work successfully.

We are very grateful to Dr.K. Vijaya Kumar, Dean-Academic, CMR College of Engineering

and Technology, for his constant support and motivation in carrying out the project work

successfully.

We are highly indebted to Dr. V A Narayana, Principal, CMR College of Engineering and

Technology, for giving permission to carry out this project in a successful and fruitful way.

We would like to thank the Teaching & Non- teaching staff of Department of Computer

Science and Engineering for their co-operation .

Finally we express our sincere thanks to Mr. Ch. Gopal Reddy, Secretary, CMR Group of

Institutions, for his continuous care. We sincerely acknowledge and thank all those who gave

support directly and indirectly in completion of this project work.

NAME ROLLNO
BANOTH GANESH (20H51A05D6)
KETHAVATH SURESH (21H55A0511)
VANAPARTHI KEERTHAN (21H55A0524)

4
 INDEX

CHAPTERS DESCRIPTION PAGE NO

Abstract i

1 Introduction 1-2

2 Background Work 3

Literature Review 4-5

Scope of Improvement 6

3 Existing System 7

4 Proposed Methodology 8-9

5 Machine Learning 10-12

Approaches
6 Platform Requirement 13

7 UML Diagrams 14-17

8 Result 18-20

Conclusion 21

References 22

5
Figures

Fig no Titles Page no

1 Upload image home screen 18
2 Select the image 18
3 Detecting image 19
4 Helmet not detected 19
5 Upload another image 19
6 Helmet detected 20
7 Detected helmet of a person 21

6
 Abstract
Motorcycle accidents have been rapidly growing through the years in many countries.
Therefore, it is necessary to develop a system for automatic detection of helmet wearing for
road safety. We now face a number of issues with India's traffic restrictions, many of which
have workable solutions. Without a helmet, it is illegal to drive a motorbike or scooter in India,
which has led to a rise in accidents and fatalities. The current system largely uses CCTV records
to monitor traffic offences, requiring traffic police to zoom in on the license plate in the event
that the rider is not wearing a helmet and look into the frame where the infringement is taking
place. However, given the high frequency of traffic offences and the rising daily use of
motorbikes, this demands a significant amount of labour and time. What if there was a system
that would check for traffic violations like riding a motorbike or moped without a helmet and,
if found, would automatically extract the license plate number of the offending car. Recent
studies have successfully completed this task using features from CNN, R-CNN, LBP, HoG,
HaaR, etc. However, these efforts have limitations in terms of effectiveness, precision, or how
quickly objects may be identified and categorized. A non-helmet rider identification system is
created in this research study in an effort to automate the process of identifying the traffic
infraction of not wearing a helmet and obtaining the license plate number of the offending
vehicle. Object Detection using Deep Learning at three layers is the key idea at play. People,
motorcycles or mopeds were discovered at the first level using YOLOv2, helmets were detected
at the second level using YOLOv3, and licence plates were detected at the final level using
YOLOv2. After that, OCR is used to retrieve the licence plate registration number (Optical
Character Recognition). All of these methods are subject to predetermined restrictions and
circumstances, particularly the component that extracts licence plate numbers. The efficiency
of the execution is essential since this job uses video as its input. We developed a
comprehensive solution for both helmet detection and licence plate number extraction using
the aforementioned approaches.

Key Words: ANPR, Machine Learning, Image Processing, OpenCV, Pytesseract , OCR.

7
CHAPTER 1

INTRODUCTION

According to transport ministry[1], in 2016 about 28 two wheeler riders die

daily because of not wearing helmet. In 2017, 31 out of 100 people died in road accidents
which shows increased rate from 21.6death per 100 accidents in 2005. Each year there are 1.4
million Traumatic Brain Injuries (TBI’s) in INDIA. About $76.5 billion dollars is spent in
treatment related to these injuries. More than 50,000 individuals die from TBI. The proposed
system aims to provide complete safety for bike riders. Recently helmets have been
made compulsory, but still, people drive without helmets. The amount of deaths has been rising
each year, especially in developing countries. Therefore, keeping public safety in mind, there
needs to be a mechanism for automatic helmet detection which can extract the number plates
of those who don’t wear helmets on roads. This sort of automation will help the administration
to issue helmet violation tickets more efficiently and ultimately aims to inhibit the violation by
two-wheeler riders.

The database consists of images of motorcycles. Existing techniques only deal

with vehicle in fixed view. However, in video, rotation is occurred and it decreases the
performance of recognition. Proposed approach where image will be stored in the form
of matrix and the output will be displayed in the form of detected numbers. The overall
work will be used for Sobel Fringe detection technique.

Two-wheeler[2] is a very popular mode of transportation in almost every country.

Observing the usefulness of helmet, Governments have made it a punishable offense to
ride a bike without helmet and have adopted manual strategies to catch the violators.
However, the existing video surveillance based methods are passive and require significant
human assistance. In general, such systems are infeasible due to involvement of humans,
whose efficiency decreases over long duration.

Automation of this process is highly desirable for reliable and robust monitoring of these
violations as well as it also significantly reduces the amount of human resources needed[3].
Also, many countries are adopting systems involving surveillance cameras at public

8
places. So, the solution for detecting violators using the existing infrastructure is also cost-
effective.

However, in order to adopt such automatic solutions certain challenges need to be addressed:

1) Real-time Implementation: Processing significant amount of information in a time

constraint manner is a challenging task. As such applications involve tasks like segmentation,
feature extraction, classification and tracking, in which a significant amount of information
need to be processed in short duration to achieve the goal of real-time implementation.[4]

2 Occlusion: In real life scenarios, the dynamic objects usually occlude each other due
to which object of interest may only be partially visible. Segmentation and classification
become difficult for these partially visible objects[5].

3) Direction of Motion: 3-dimensional objects in general have different appearance from

different angles. It is well known that accuracy of classifiers depends on features used
which in turn depends on angle to some extent. A reasonable example is to consider
appearance of a bike rider from front view and side view.

4) Temporal Changes in Conditions: Over time, there are many changes in environment
conditions such as illumination, shadows, etc. There may be subtle or immediate changes
which increase complexity of tasks like background modelling[6].

5) Quality of Video Feed: Generally, CCTV cameras capture low resolution video. Also,
conditions such as low light, bad weather complicate it further. Due to such limitations, tasks
such as segmentation, classification and tracking become even more difficult. As stated
in, successful framework for surveillance application should have useful properties such
as real-time performance, fine tuning, robust to sudden changes and predictive. Keeping
these challenges and desired properties in mind, we propose a method for automatic
detection of bike-riders without helmet using feed from existing security cameras, which works
in real time[7].

9
CHAPTER 2

BACKGROUND WORK

Literature Review
1. Machine Vision Techniques for Motorcycle Safety Helmet Detection

This paper presents a system which automatically detects motorcycle riders and
determines that they are wearing safety helmets or not. The system extracts moving objects
and classifies them as a motorcycle or other moving objects based on features extracted from
their region properties using K-Nearest Neighbor / (KNN) classifier. The heads of the riders
on the recognized motorcycle are then counted and segmented based on projection
profiling. The system classifies the head as wearing a helmet or not using KNN based on
features derived from 4 sections of the segmented head region. Experiment results show an
average correct detection rate for near lane, far lane, and both lanes as 84%, 68%, and 74%,
respectively[8].

2. . Smart Helmet Using GSM and GPS Technology

In this, Author has discussed safety and security of the bikers against road accident.
Smart helmet has special idea which makes motorcycle driving safety than before, this is
implemented using GSM and GPs technology. Other advantages of this project is to measure
the alcohol level of the drunken people who is riding the bike. Whenever the alcohol level
crosses the predefined value, the alarm starts and get notification about the drunken driver. The
author have also discussed about the accident detector and the sensor will active the GPS and
find the location and further SMS will send to ambulance or family members.[9]

3. . SMART HELMET– A Review Paper:

According to the recent Research paper in 2016 titled ‘2 Helmet using GSM and GPS
technology for accident detection and reporting system’, The author specially developed this
project to improve the safety of the bikers. The objective of this project is to study and
understand the concept of RF transmitter and RF receiver circuit. The project uses ARM7,
GSM and GPS module. The project also uses buzzer for indication purpose. Whenever
the accident will occur then accident spot will be note down and information will send out on
the registered mobile number. The major disadvantage of this project is they are not using any

10
display device for showing the current status. Also the cost of helmet is still high since helmet
is designed for only one purpose. According to the Research paper in 2015 titled
‘Microcontroller based smart wear for driver safety’, In this paper author has discussed on the
speed of the vehicle. In this application the project will be monitoring the areas in which the
vehicle will be passing. On entering any cautionary areas like schools, hospitals, etc the
speed of the vehicle will be controlled to a predefined limit. LCD is used for showing the
various types of messages after wearing the helmet. The author has worked only on the
phenomenon of accident which is generally happens due to drunk and drive. But as we know
that the accident in the area is not happens only due to consuming alcohol but also other
parameters like speed are also responsible[9].

4. Literature Survey On Image Filtering Techniques

Image processing has become a common technique for making images more

comprehensible to the human eye. Images acquired are found to be corrupted with noise

in many cases. There are many methods available to remove impulse noise in gray scale

and color images. But very little has been done for the removal of additive noise in color

images. Of the many filters presented, most of them are only for gray scale images. The filtering

techniques developed for gray scale images can be extended to color images by applying

it to the different color components separately but it is also evident that they can partially

destroy image details. The existing systems includes Conservative Smoothing, linear filters,

non-linear filters like median filter and fuzzy filter, adaptive filter, wavelet based filter etc.

These techniques have a number of advantages and also disadvantages[3].

5. Digital Image Processing Techniques in Character Recognition of Indian

Languages.

This paper presents a brief overview of digital image processing techniques such as
Feature Extraction, Image Restoration and Image Enhancement. A brief history of OCR and
various approaches to character recognition is also discussed in this paper. Handwritten
character recognition is always a frontier area of research in the field of pattern recognition.
There is a large demand for OCR on hand written documents in Image processing. Even

11
though, sufficient studies have performed in foreign scripts like Arabic, Chinese and
Japanese, only a very few work can be traced for handwritten character recognition mainly
for the south Indian scripts. OCR system development for Indian script has many application
areas like preserving manuscripts and ancient literatures written in different Indian scripts and
making digital libraries for the documents. Feature extraction and classification are essential
steps of character recognition process affecting the overall accuracy of the recognition
system[10].

6. Limitations and challenges in Existing Systems:

Bikers do not wear helmets in the region where traffic checking is not done. There is
a tendency of the driver to wear helmet only where the anticipate checking may takes
place, they do not wear helmet where no checking is done. The vehicle may be turn on or
may be stolen by passing the ignition switch.

Testing alcohol content present in blood in each individual rider in big countries like
India is almost impossible. Accidents due phone calls as previous helmets do not contain
Bluetooth speakers.

7. Automatic detection of motorcyclists without the helmet.

This paper aims to explain and illustrate an automatic method for motorcycles
detection and classification of public roads and a system for automatic detection of
motorcyclists without the helmet. For this, a hybrid descriptor for features extraction is
proposed based on Local Binary Pattern, Histograms of Oriented Gradients and the Hough
Transform descriptors. Traffic images captured by cameras were used. The best result obtained
from classification was an accuracy rate of 0.9767, and the best result obtained from the helmet,
detection was an accuracy rate of 0.9423[4].

8. Motorcycle Detection and Tracking System with Occlusion Segmentation.

The method uses the visual length, visual width. and Pixel Ratio to detect the
classes of the motorcycle occlusions and segment the motorcycle from each occlusive class.

12
Because the motorcycle riders must put on their helmets, the helmet detection or search method
is used to make sure whether the helmet/motorcycle exits or not.

Experiments obtained by using complex road scenes are reported, which demonstrate
the validity of the method in terms of robustness, accuracy, and timely responses.

9. Automatic Detection of Bike-riders without Helmet using Surveillance Videos in

Real-time.

In this paper, we propose an approach for automatic detection of bike-riders

without helmet using surveillance videos in real time. The proposed approach first detects

bike riders from surveillance video using background subtraction and object segmentation.

Then it determines whether bike-rider is using a helmet or not using visual features and binary
classifier. Also, we present a consolidation approach for violation reporting which helps
in improving the reliability of the proposed approach. In order to evaluate our approach, we
have provided a performance comparison of three widely used feature representations namely
histogram of oriented gradients (HOG), scale-invariant feature transform (SIFT), and local
binary patterns (LBP) for classification[11].

Scope of Improvement

Any system always has a scope for improvements and more advancement. All the
systems studied under the literature survey have their own different features. All the systems
proposed till date are used only for sending message in case of accident. There could be such a
system where only alcohol detection is present. Here in this system many advanced features
are added and also the previous features are clubbed in a single system. It will send message
automatically when rider met an accident with helmet on. RF transmitter and receiver are used
for starting the two wheeler, if rider not wearing the helmet the bike will not get start. The
alcohol sensor will sense the alcohol and it will lock the ignition if drunk. The solar sense is
generating power for the system. It can tracked easily with location when bike is stolen.It can
also use to receive call while driving through wireless Bluetooth Speakers.

13
 CHAPTER 3

Existing System

The system proposed isolates the bikes from the images by approximation crops the
most probable area where helmet might be present and then feeds it to the feature abstraction
and matching system. Automatic detection of bike-riders without helmet falls under broad
category of anomaly detection in surveillance videos. As explained in, effective automatic
surveillance system generally involve following tasks: environment modeling, detection,
tracking and classification of moving objects. It uses circle arc detection method based on the
Hough transform. The major limitation of this approach is that it tries to locate helmet in
the full frame which is computationally expensive and also it may often confuse other similar
shaped objects as helmet.

Existing System

14
CHAPTER 4

Proposed Methodology

1.Flowchart of the proposed methodology

2. Moving object detection:

The first task in helmet identification is to detect a moving vehicle. It is the first step
before performing more sophisticated functions such as tracking or categorization of
vehicles. Rather than immediately processing the entire video, the example starts by obtaining
an initial video frame in which the moving objects are segmented from the background.
Processing only the initial few frames helps to take the steps required to process the video. The
foreground detector needs a certain number of video frames to initialize the Gaussian mixture
model. The foreground segmentation process is not perfect and often includes undesirable
noise. Next, we will find bounding boxes of each connected component corresponding
to a moving vehicle. Generally, more than one blob is detected apart from moving vehicles
such as pedestrians, trees, dogs and other small noises. All the blobs that consist of less than
n number of pixels are discarded (in our case n is 150 pixels). This way, we only remain with
the moving vehicle. But there are a lot of gaps in the blob, that is, it is not one coherent blob.
We use the morphological opening to remove the noise and to fill gaps in the detected objects
which makes the blob more coherent. Once the blob is found, the raw image is extracted that
is hidden behind the blob.

15
3. Vehicle classification:
The next step is to classify the moving vehicle extracted in the last part. To classify
vehicle, we are going to use the number of machine learning algorithms, from classical machine
learning algorithms to modern deep neural networks, to see which approach works best in
vehicle classification with limited data. A vehicle can be classified into two categories two-
wheelers or four-wheelers. We are only interested in two wheelers Figure 1 since we want to
detect the presence of a helmet. The system proceeds further only if a two-wheeler is detected.
Else, it discards this vehicle and looks for other vehicles and the cycle continues. We will
collect the training data required for the classification of a vehicle on our own. We will
capture the images of various vehicles in various positions. Almost same number of images,
1000, were gathered for both the classes two-wheelers or four-wheelers. If there are equal
number of training images from both classes then it eliminates the problem of class imbalance
and leads to better performance of the classifier. The training images contain a vehicle
surrounded by other objects of interest such as trees, footpath, how a vehicle is normally seen
on roads. Although this dataset is not the most representative of the real world moving objects,
it is still enough to train and test the effectiveness of various machine learning algorithms to
check the feasibility of the approach. The images will be converted to grayscale. Raw pixel
values were fed to the classifier.

4. Helmet detection:
Using the same approach as applied to identify the type of vehicle, we detect
whether the rider is wearing a helmet. The images that will be used to train a helmet detector
will be cropped version of the two-wheeler images focusing on the head region of the rider.
Using this technique, we will be able to maintain the class balance, that is, there will be same
number of images where the rider was wearing a helmet and where the rider was not wearing
a helmet. We used numerous machine learning classifiers in order to select the best one for this
task.

5. License plate extraction:

After the previous steps, in case if the rider of a two-wheeler is not wearing a helmet,
our next step is to extract the license plate of the vehicle. We extract the region of interest
from our cropped image by giving the appropriate coordinates.

16
 CHAPTER 5

MACHINE LEARNING APPROACHES

A total of five machine learning classifiers will be used to test which one performs better in our

scenario. The classifiers are Random Forest:

This algorithm is based on decision trees. Here, instead of building one tree, a lot of
trees are grown in parallel. All these trees are fed only a subset of data points and a subset
of features. The sub setting ensures diversity among the trees. After training, each tree votes
for a class and a final class is chosen based on the highest number of votes.

Gradient Boosted Trees:

This algorithm is also based on decision trees like the random forest. However, instead
of constructing a lot of trees in parallel, trees are constructed sequentially one after the other.
Each tree improves the loss by rectifying the error made by the previous tree while training.

Support Vector Machine :

An SVM [9] creates a hyperplane (a plane in n-dimensions) which divides all the classes
in the training data from one another in such a way that the difference between the two classes
is maximum. This algorithm takes a lot of computing resources to complete the training as
compared to the aforementioned classification techniques.

Deep Neural Networks :

Deep neural networks are an improvement over conventional neural networks.

These are neural networks with a large number of layers where each layer has a plethora of
nodes. Deep learning is being used to achieve state-of-the-art results in the field of computer
vision and natural language processing. Deep neural networks require a lot of training data as
compared to conventional machine learning algorithms to outperform them. And that’s why
we were interested to see how well would this technique work in case of a small training data.
If we had a large training data (hundreds of thousands of images for each class), the choice
obviously would have been a deep neural network. However, the small number of images

17
becomes a bottleneck for such networks. We used a 10-layer network with 50 nodes in each
layer.

TensorFlow:

The deep learning library in the world is Google's TensorFlow. Google product uses machine
learning in all of its products to improve the search engine, translation, image captioning or
recommendations. example, Google users can experience a faster and more refined the search with AI.
If the user types a keyword the search bar, Google provides a recommendation about what could be
the next word.

Google wants to use machine learning to take advantage of their massive datasets to
give users the best experience. Three different groups use machine learning:

 Researchers

 Data scientists

 Programmers.

They can all use the same toolset to collaborate with each other and improve their efficiency.

Google does not just have any data; they have the world's most massive computer, so Tensor
Flow was built to scale. TensorFlow is a library developed by the Google Brain Team to
accelerate machine learning and deep neural network research.

It was built to run on multiple CPUs or GPUs and even mobile operating systems, and it has
several wrappers in several languages like Python, C++ or Java.

TensorFlow architecture works in three parts:

• Preprocessing the data

• Build the model

• Train and estimate the model

It is called TensorFlow because it takes input as a multi-dimensional array, also known as

tensors. You can construct a sort of flowchart of operations (called a Graph) that you want to

18
perform on that input. The input goes in at one end, and then it flows through this system of
multiple operations and comes out the other end as output.

OpenCV:

OpenCV is the leading open source library for computer vision, image processing and machine
learning, and now features GPU acceleration for real-time operation.

OpenCV is released under a BSD license and hence it’s free for both academic and
commercial use. It has C++, C, Python and Java interfaces and supports Windows, Linux,
Mac OS, iOS and Android. OpenCV was designed for computational efficiency and with a
strong focus on real-time applications. Written in optimized C/C++, the library can take
advantage of multi-core processing. Adopted all around the world, OpenCV has more than
47 thousand people of user community and estimated number of downloads exceeding 6
million.

Usage ranges from interactive art, to mines inspection, stitching maps on the web or
through advanced robotics.

OpenCV is being used for a very wide range of applications which include:

 Street view image stitching

 Automated inspection and surveillance

 Robot and driver-less car navigation and control

 Medical image analysis

 Video/image search and retrieval

 Movies - 3D structure from motion

 Interactive art installations

19
CHAPTER 6

Platform Requirement
1.Software Requirement

Machine Learning Libraries: TensorFlow, OpenCV

Scripting language: Python

Database: MongoDB / Cloud storage (AWS)

Platform: Linux, Windows

API tools: Coco API / YOLOv2 API

2. Hardware Requirement

Hardware: 8GB RAM, 6GB Graphics, i7 Processor

Graphics Driver: NVDIA GTX 1050

20
CHAPTER 7

UML Diagrams

1. System Architecture

21
2. Use case Diagram

3.Component Diagram

22
4.State Chart diagram

5.Activity Diagram

23
6.Deployment Diagram

24
CHAPTER 8

RESULT

After setting path double click on ‘run.bat’ file to run project and to get below screen

Fig1. Upload image home screen

In above screen click on ‘Upload Image’ button and upload image

Fig2.Select the image

I choose the picture labelled "5.png" in the screen above, then I clicked the "Open" button to
load the image. To determine if a picture includes a person and a motorbike, click the 'Detect
Motor Bike & Person' button now.

25
 Fig3.Detecting image

Click the "Detect Helmet" button to determine if the individual riding the bike in the above
screen is wearing a helmet or not.

Fig4.Helmet not detected

The programme on the top screen saw that the individual wasn't wearing a helmet, so it pulled
the person's number from the car and displayed it next to the text space. We will now verify
using a helmet picture.

Fig5.Upload another image

I've uploaded image 4.png to the screen above; when you click the "Detect Motor Bike &
Person" button, the following result will appear.

26
 Fig6.Helmet detected

Clicking the "Detect Helmet" button after Yolo identified a rider on a motorcycle in the
previous creen produced the results shown below.

Fig7.Detected helmet of a person

The programme on the above screen saw that the guy was wearing a helmet and that the label
was visible around his head. The application stopped there without scanning the licence plate.

We have trained a few photographs to extract number plates for this project, but if you want to
extract for new images, email them to us so we may incorporate them in the yolo model to
extract number plates for new images as well.

27
 Conclusion
A video file is used as the input for a system being built to detect non-helmeted riders. The
motorcycle's licence plate number is retrieved and shown if the rider in the video clip is not
wearing a helmet while operating the machine. Motorcycle, person, helmet, and licence plate
detection all employ the object detection concept with the YOLO architecture. In cases when
the rider is not wearing a helmet, OCR is utilised to retrieve the licence plate number. In order
to be utilised for various purposes, not only the characters but also the frame from which they
are retrieved are removed. The project's goals have all been properly met.

28
 References

1. J.Chiverton, “Helmet Presence Classification with Motorcycle Detection And Tracking”,IET

Intelligent Transport Systems,Vol. 6, Issue 3, pp. 259–269, March 2012.

2. Rattapoom Waranusast, Nannaphat Bundon, Vasan Timtong and Chainarong Tangnoi,

“Machine Vision techniques for Motorcycle Safety Helmet Detection”, 28th International
Conference on Image and Vision Computing New Zealand, pp 35-40, IVCNZ 2013.

3. Romuere Silva, Kelson Aires, Thiago Santos, Kalyf Abdala, Rodrigo Veras, Andr´e Soares,
“Automatic Detection Of Motorcyclists without Helmet”, 2013 XXXIX Latin America
Computing Conference (CLEI).IEEE,2013.

4. Romuere Silva, “Helmet Detection on Motorcyclists Using Image Descriptors and

Classifiers”, 27th SIBGRAPI Conference on Graphics, Patterns and Images. IEEE, 2014.

5. Thepnimit Marayatr, Pinit Kumhom, “Motorcyclist‟s Helmet Wearing Detection Using

Image Processing”, Advanced Materials Research Vol 931- 932,pp. 588-592,May-2014.

6. Amir Mukhtar, Tong Boon Tang, “Vision Based Motorcycle Detection using HOG features”,
IEEE International Conference on Signal and Image Processing Applications (ICSIPA).IEEE,
2015.

7. Abu H. M. Rubaiyat, Tanjin T. Toma, Masoumeh Kalantari-Khandani, “Automatic

Detection of Helmet Uses for Construction Safety”, IEEE/WIC/ACM International Conference
on Web Intelligence Workshops(WIW).IEEE, 2016.

8. XINHUA JIANG “A Study of Low-resolution Safety Helmet Image Recognition Combining

Statistical Features with Artificial Neural Network”.ISSN: 1473-804x

29